Search

Copyright of GeSec: Revista de Gestao e Secretariado is the property of Sindicato das Secretarias e Secretarios do Estado de Sao Paulo (SINSESP) and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

O presente trabalho tem por objetivo realizar a previsão das séries temporais do Índice Bovespa (Ibovespa) utilizando os métodos de Redes Neurais Artificiais – RNA e Adaptive Neuro Fuzzy Inference Systems – ANFIS, tendo em vista a busca de alternativas a modelos lineares que podem ignorar certos aspectos das estruturas dinâmicas existentes no mercado de ações. A metodologia foi de caráter descritiva e quantitativa. Para as técnicas de análise de dados, tendo em vista as possíveis características das séries temporais financeiras, optou-se por utilizar dois métodos não lineares para a análise, ou seja, uma rede neural artificial (RNA) com arquitetura MLP (multilayer perceptron) com algoritmo Levenberg-Marquardt e também o método Adaptative Neuro-Fuzzy Inference System (ANFIS). Verificou-se um bom ajuste do modelo para a série estudada, o que confirma que modelos híbridos utilizando Sistemas Fuzzy (em termos de regras difusas) apresentam características que podem ser úteis para lidar com séries temporais que possuem estruturas dinâmicas complexas e não-lineares como é o caso das séries temporais financeiras. Os resultados encontrados demonstram que por meio de métodos como o Adaptive Neuro Fuzzy Inference Systems – ANFIS, pode-se realizar previsões para o mercado acionário com certa previsibilidade, comprovando que eventos ocorridos no passado podem vir a determinar eventos futuros como salienta a teoria de Charles Dow e refutando a hipótese do Random Walk de Bachelier (1964) para a presente amostra.

Copyright of Revista Foco (Interdisciplinary Studies Journal) is the property of Revista Foco and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of GeSec: Revista de Gestao e Secretariado is the property of Sindicato das Secretarias e Secretarios do Estado de Sao Paulo (SINSESP) and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Environmental & Social Management Journal / Revista de Gestão Social e Ambiental is the property of Environmental & Social Management Journal and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of GeSec: Revista de Gestao e Secretariado is the property of Sindicato das Secretarias e Secretarios do Estado de Sao Paulo (SINSESP) and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

O algoritmo de redes neurais artificiais (RNA), são modelos computacionais que possuem a capacidade de realizar generalização, inferências, identificação e classificação de informações e padrões. De modo que, nesse trabalho foi desenvolvido um estudo através da criação de uma rede neural classificadora de padrões a fim de identificar e classificar os tipos de curto-circuito que ocorrem no sistema elétrico de distribuição. Assim, desenvolveu-se uma rede neural perceptron multicamadas no software Matlab com 3 camadas escondidas, 25 neurônios em cada camada escondida e função de ativação do tipo tangente hiperbólica. A PMC foi treinada utilizando-se dados simulados de curto-circuito no software ATPDraw e apresentou uma eficiência de 94.7% na identificação dos curtos-circuitos na etapa de validação. A rede treinada também conseguiu avaliar curtos-circuitos em um barramento de testes de 9 barras da IEEE demonstrando potencial para ser aplicada como medida adicional de informação da rede nos centros de operação integrado (COI).

Copyright of Estudios Gerenciales is the property of Universidad ICESI and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Vulnerabilidades em Java correspondem a 91% de todos as cyber-infecções monitoradas na rede mundial de computadores. O presente trabalho cria um Antivírus da Próxima Geração, dotado de aprendizado estatístico e Inteligência Artificial, especializado na detecção de malwares Java. O antivírus proposto monitora e pondera, estatisticamente, 6.824 ações que os arquivos suspeitos possam fazer quando executados. O antivírus proposto alcança um desempenho médio de 91,58% na distinção entre arquivos benignos e malwares. No lugar de modelos baseados em listas negras, o antivírus proposto permite a detecção de malwares de forma preventiva, e não de maneira reativa, como acontece com antivírus comerciais.

The cultivation of soybean (Glycine max (L.) Merr.) stands out in the national and world scenario. Brazil, from different points of view, has a high productive capacity of this oilseed, which resulted in the classification of the country as the world's largest producer. The state of Minas Gerais is one of Brazil's agricultural granaries, and in soybean culture it is the sixth largest producing state. Thus, the objective was to evaluate the interaction genotypes x environments in soybean genotypes grown in Alto Paranaíba, Minas Gerais, and to identify genotypes with general response, those recommended for favorable or unfavorable environments through the different methodologies of adaptability and stability. Five experiments were conducted in the experimental block design, with four repetitions, in the 2020/2021 agricultural season, in which 14 soybean genotypes were analyzed by productivity, in kg ha-1. An individual analysis of variance was performed (for each environment) along with joint analysis of the experiments, average grouping test (for each environment), decomposition of the interaction between genotypes and pairs of environments and the adaptability and stability analyses were performed. It was concluded that the interaction genotypes x environments was significant and between pairs of environments it was predominantly complex, the genotypes M5917IPRO, TMG7063IPRO and BMXFocoIPRO are recommended for general environment, the genotypes TMG2374IPRO and TMG7363RR are recommended for favorable environments, and the genotypes BS2606IPRO, TEC7849IPRO and BRS5980IPRO are recommended for unfavorable environments. [ABSTRACT FROM AUTHOR]

As redes neurais artificiais (RNA), são modelos computacionais inspiradas na forma de trabalho do sistema nervoso dos seres vivos, esses modelos podem ser utilizados para processamento e classificação de dados e aplicações, tais como previsão de séries e funções. Dessa forma, neste trabalho utilizou-se uma rede neural com retardo de tempo (TDNN) para prever a demanda de energia ativa no barramento P4 na cidade de Presidente Prudente.

Copyright of Tecno-Lógica is the property of Tecno-Logica and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Resumo Estudos comprovam que a temperatura da superfície do mar nos oceanos tropicais Pacífico e Atlântico é a principal variável física influenciadora das condições climáticas em várias áreas do globo. Assim, as observações e previsões de variáveis oceânicas, como a temperatura da superfície do mar, constituem-se uma ferramenta valiosa para o monitoramento do clima e para um melhor aproveitamento dos recursos naturais, principalmente em regiões que são vulneráveis aos impactos causados pelas mudanças climáticas, como é o caso do Nordeste do Brasil. O presente trabalho tem por objetivo aplicar e avaliar um método que visa utilizar as Redes Neurais Artificiais para prever a temperatura da superfície do mar no oceano Atlântico Tropical e o Gradiente meridional de anomalias de temperatura da superfície do mar no Atlântico Tropical. Os resultados mostraram que as correlações, significativas a 95% segundo o teste-t de Student, foram melhores para os trimestres FMA e MAM, onde os coeficientes apresentaram valores superiores a 0,75 para algumas regiões do ATN. Para as estimativas do Gradiente Meridional, a RNA apresentou desempenho similar à persistência. Os melhores resultados ocorrem ao se usar a média da TSM do trimestre DJF. Para essa configuração o coeficiente de correlação foi 0,74.

Resumo O preenchimento de falhas em séries temporais de precipitação é um importante processo para aplicações em hidrologia, visando o aproveitamento de longas séries, evitando que as mesmas sejam descartadas. Desse modo, este estudo teve como objetivo realizar o preenchimento de falhas em séries históricas de precipitação diária no Rio Grande do Sul (RS), auxiliando no aproveitamento desses dados em estudos que necessitem de analises de longo prazo. Para tanto, foram utilizadas séries históricas de 287 estações, no período entre 1987 e 2016 e aplicados os métodos de Regressão Linear Múltipla (RLM) e Redes Neurais Artificiais (RNA), comparando e avaliando os valores preenchidos. Um algoritmo foi desenvolvido para realizar as seguintes operações: i) identificar os dias com falhas em cada estação; ii) identificar as estações que podem ser utilizadas para o preenchimento de cada falha; iii) identificar todas as combinações de entrada para o preenchimento de falhas em cada estação; iv) realizar o ajuste/treinamento dos modelos RLM e RNA; v) realizar a validação dos modelos com base no período sem falhas de cada estação. Os principais resultados indicam que a maior densidade de estações pluviométricas favorece o processo de preenchimento de falhas em séries históricas de precipitação, melhorando a qualidade da série preenchida. O preenchimento de falhas apresentou maior coeficiente de determinação e menor erro médio absoluto usando o modelo RLM em relação às RNA, possivelmente em função da forte correlação linear entre os dados de precipitação de cada local em relação a sua vizinhança. O modelo de RLM apresentou um coeficiente de determinação (R2) médio de 0,697, enquanto que o modelo RNA obteve media de 0,675. Levando-se em conta a análise por meio do erro médio absoluto (EMA), a média dos valores foi de 2,27 mm para a RLM, enquanto para a RNA o erro ficou em 2,31 mm. Conclui-se, considerando o conjunto de dados de precipitação diária do RS, que houve uma pequena superioridade do método RLM em relação à RNA.

Esse artigo trabalha com a possibilidade de identificar e unificar, automaticamente, volumes significativos de demandas judiciais em tramitação que possuam o mesmo fato e tese jurídica. Com a identificação e a unificação dos processos em agrupamentos, objetiva-se criar pendências no Sistema de Processo Eletrônico com a finalidade de informar a possibilidade de ocorrência de conexão às diferentes unidades judiciais que receberam as causas por distribuição, alertando e facilitando a análise pelo Julgador. São aplicadas técnicas de Processamento de Linguagem Natural, aprendizagem por similaridade e Redes Neurais Artificiais. A solução de Inteligência Artificial (IA) construída, chamada Berna, encontra-se em produção no Poder Judiciário Goiano. A precisão de 96% nos estudos de casos demonstra a efetividade do método.

Os principais defeitos em para-raios estão comumente associados a elevações de corrente e, por isso, ao aquecimento por efeito Joule. Assim, a termovisão nestes equipamentos é uma técnica adequada para a detecção de defeitos. Porém, tal uso da termovisão depende de um operador experiente para interpretar os resultados obtidos, sujeitando o processo a erros de interpretação. Diante disso, este trabalho tem como objetivo aplicar técnicas de Inteligência Artificial na classificação de imagens térmicas de para-raios como forma de tornar o processo mais confiável. Foram desenvolvidos algoritmos de inteligência artificial baseados em técnicas de Redes Neurais Artificiais e deep learning, tendo em vista que muitos autores obtiveram sucesso na utilização destes métodos no diagnóstico de falhas em outros equipamentos do sistema elétrico de potência. Os resultados obtidos mostraram que as redes neurais desenvolvidas pelo método backpropagation apresentaram uma boa eficiência ao classificar imagens de para-raios e que não há necessidade de segmentar o para-raios da imagem termográfica para realizar a classificação.

Copyright of Revista Ambiente e Água is the property of Revista Ambiente e Agua and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

A Olimpíada Brasileira de Robótica é uma competição entre estudantes, na qual um dos desafios é que os robôs capturem esferas dispostas numa sala. Este trabalho propõe uma solução para esse problema utilizando visão computacional e redes neurais artificiais, as quais fornecem como saída final um vetor contendo a localização do ponto do centro da esfera mais próxima e seu raio na imagem. Isso permite que os pontos que delimitam a esfera sejam determinados. O modelo obtido ao final do trabalho em um computador de uso geral foi satisfatório, obtendo uma acurácia de detecção de aproximadamente 90%. Para seu desenvolvimento, foi utilizado um banco de dados com 1440 imagens de situações possíveis com as quais o robô pode se deparar. A rede já foi testada numa possível placa a ser embarcada, porém resta ainda fazer os testes em um robô real.

Currently, in almost all segments of the production chain, automation is a requirement for productivity improvement. With respect to nuclear facilities, active online monitoring is one of best practices for nuclear security and safety maintenance, to prevent incidents that could compromise a particular installation. In this context, spectral signature monitoring automation can be explored, aiming at the rapid identification of adverse events, such as radiological accidents. The main objective of this work was an automated radionuclides classification technique establishment, using an Artificial Neural Networks (ANN) architecture. The methodology used consisted basically of simulating the geometry of an established experimental apparatus, using the MCNP5 code, obtaining the simulated gamma spectral signature for the studied nuclides. The simulated spectra were used to compose the ANN training and testing data set, while the experimental spectra were subjected to the artificial intelligence model classification, in order to allow the neural network quality assessment. The final developed architecture of ANN was correct to recognize the experimental spectra of 60Co, 137Cs and 152Eu. Therefore, the results were satisfactory and proved automation technique development viable.

Background: Wastewater from the mineral fertilizer production, agribusiness containing ammonium ions causes significant harm to fish farming; therefore, it must be purified before discharge. Ion-exchange sorption is a promising method for isolating ammonium cations. The object of the study was a chemisorption fiber VION KN-1, which has developed surface and high sorption rate. Purpose: To study the sorption kinetics of ammonium cations from aqueous solutions on VION KN-1; to train an ANN to predict the degree of recovery of ammonium ions from wastewater using Statistica Neural Networks Version 13. Methods: The ammonium ion concentration in the solution was established by direct potentiometry. Sorption isotherms were constructed using the method of variable concentrations. To determine the limiting stage, the obtained kinetic dependencies were represented in the coordinates of the Boyd-Adamson equations for internal/external diffusion. Results and Discussion: During sorption from solutions with different ammonium nitrogen contents, the values of distribution coefficients (Kd) are at the level of 2.3ꞏ10³ cm³/g, which significantly exceeds this parameter for granular ionites. Experimental sorption data were verified using Freundlich (R² = 0.9224) and Langmuir (R² = 0.9996) isotherms. The maximum degree of recovery (over 96 %) was achieved by passing a solution with a concentration of 11.3 mmol/dm³. Using an array of experimental data, the MLP-3-5-1 neural network was trained. The coefficient of determination R² = 0.999420 obtained for the training sample characterizes high network performance. Conclusions: The Langmuir equation better describes the process of NH4+ sorption on a fibrous sorbent. It is reasonable to use VION KN-1 at the fine treatment stage. Ammonium ion desorption from the fiber was performed by acid solution. The resulting solutions of ammonium salts can be used as liquid fertilizers. The trained neural networks can be used to predict the degree of recovery of ammonium ions by sorbent VION KN-1. [ABSTRACT FROM AUTHOR]

Copyright of Revista Gestão & Tecnologia is the property of Revista Gestao & Tecnologia and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Objetivou-se com este trabalho definir configurações acuradas de redes neurais artificiais (RNA) para estimar o afilamento do fuste de indivíduos de eucalipto com seis anos de idade. Foi realizada uma cubagem rigorosa em um povoamento comercial no município de Paragominas, mesorregião sudeste do Estado do Pará. Para maior precisão de cubagem, todos os clones foram separados em três classes diamétricas, sendo abatidas cinco árvores por classe, totalizando 60 árvores abatidas. Para o banco de dados foram treinadas 240 RNA no software Neuro versão 4.06. As RNA treinadas foram do tipo Multilayer Perceptron (MLP), com o algoritmo de aprendizado Resilient Propagation RPROP+, com diferentes funções de ativação e arquitetura, sendo estas avaliadas quanto o bias, raiz quadrada do erro médio, variância, erro padrão da estimativa e coeficiente de correlação. As RNA com menor valor ponderado foram as 165, 202, 204, 203 e 177, as quais apresentaram função de ativação do tipo sigmoidal. O coeficiente de correlação apresentou valores maiores que 0,99 para o treinamento e 0,98 para a validação das RNA, nas RNA treinadas. As RNA não foram tendenciosas e possuem capacidade de estimar o taper do eucalipto com acurácia. A RNA 165, com arquitetura de 5-7-1 e função de ativação sigmoidal, foi a que apresentou melhores resultados.

Resumo A quantidade de dados não estruturados cresce com a popularização da Internet. Textos em linguagem natural representam um conjunto relevante e significativo para análise e produção de conhecimento. Este trabalho propõe uma análise quantitativa das etapas de pré-processamento e de treinamento de um classificador de textos, que utiliza os sentimentos expressos pelos usuários como atributo. Para realização dos experimentos, foram utilizadas Redes Neurais Artificiais, como algoritmo classificador, e textos provenientes dos sites Amazon, IMDB e Yelp. A base textual permite análise da expressão de sentimentos positivos e negativos dos usuários em avaliações de produtos e serviços em textos não estruturados. Foram realizados dois processos distintos de pré-processamento e diferentes treinamentos das Redes Neurais Artificiais para classificação do conjunto textual. Os resultados confirmam, quantitativamente, a importância das etapas de pré-processamento e de treinamento do classificador, evidenciando a importância do vocabulário selecionado para a representação do texto e para a classificação. As técnicas de classificação disponíveis alcançam resultados satisfatórios. No entanto, mesmo utilizando-se dois processos distintos de pré-processamento e identificando-se o melhor processo de treinamento, não foi possível eliminar, totalmente, as dificuldades de aprendizado e compreensão do modelo para as classificações de sentimentos que envolviam características subjetivas da expressão do sentimento humano.

Supernovas são eventos catastróficos no qual algumas estrelas explodem. A classificação de supernovas é feita por especialistas por meio da análise dos espectros de luz que apresentam linhas de absorção e emissão em determinadas regiões do comprimento de onda. Os espectros de luz das supernovas apresentam padrões que podem ser usados em algoritmos de aprendizagem de máquina possibilitando assim a classificação automática e inteligente das supernovas. A classificação automática é essencial para o processamento de grande quantidade de dados em equipamentos instalados em lugares remotos, onde não é sempre possível a presença de um especialista. O objetivo deste trabalho é apresentar a CINTIA 2, aprimoramento do Classificador Inteligente de supernovas do tipo Ia, que usa uma hierarquia de redes neurais binárias do tipo Perceptron para classificar as supernovas nos tipos Ia, Ib, Ic e II. Apresentamos a arquitetura da CINTIA 2 e a ferramenta daí proveniente, desenvolvida nas linguagens de programação Python e C++. Os resultados obtidos apresentam ótimo desempenho, principalmente na classificação dos tipos Ia e II. Uma comparação com trabalhos encontrados na literatura mostra que a CINTIA 2 é superior em quantidade e diversidade de dados e alcança índices de classificação equiparáveis aos demais classificadores.

Copyright of Estudios Gerenciales is the property of Universidad ICESI and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Neste trabalho foi proposta a utilização de Redes Neurais Artificiais (RNAs) com o Método de Levenberg-Marquardt (MLM) para estimar o fluxo de permeado de uma bebida à base de açaí através do processo de microfiltração tangencial com duas membranas cerâmicas com poros de diâmetro médio de 0,8 e 1,2 𝜇m em escoamento turbulento. As variáveis de entrada das RNAs foram o número de Reynolds, a pressão transmembrana e o tempo de microfiltração; as variáveis de saída foram os correspondentes valores de fluxo de permeado. Os resultados fornecidos pelas RNAs tiveram baixos erros médios, sendo 4,1%, 10,6% e 7,6% para a membrana de 0,8 𝜇m e 5,3%, 4,7% e 9,9%, para a membrana de 1,2 𝜇m, para os conjuntos de treinamento, validação e teste, respectivamente.

Today’s world is going through what is known as the Fourth Industrial Revolution. Robots have been gaining more and more space in the industry and going beyond expectations. The use of robots in industry is related to the increasing production and the quality of the electronic products. For an accurate movement of a robot manipulator it is necessary to obtain its inverse kinematic model, however, obtaining this model requires the challenging solution of a set of nonlinear equations. For this system of nonlinear equations there is no generic solution method. In view of this matter, this research aims to solve the problem of the inverse kinematics of a robot manipulator using artificial neural networks without the need to model the robot’s direct kinematics. Two neural network training methodologies, called offline training and online training were used. The basic difference between these two is that in the offline methodology all training points are obtained before any training of the neural network occurs, whereas in online training the use of a training method is recurrent. As the robot moves, new training points are obtained and training processes with the new acquired points are executed, allowing a learning process of continuous inverse kinematics. To validate the proposed methodology a prototype of a manipulated planar robot with one degree of freedom was developed and several architectures of neural networks were tested to find the optimal architecture. The offline training methodology obtained very satisfactory results for most of the neural network architectures tested. The online training only achieved satisfactory results in neural network architectures with quantities of neurons much larger than the quantities used in the architectures used in offline training and still obtained inferior results. The neural networks trained in offline mode, when compared to the training networks in the online mode, presented a greater capacity of generalization and a smaller value of output error. The online training has only achieved satisfactory results in neural network architectures with quantities of neurons much larger than the quantities used in the trained architectures in offline mode.

Copyright of Revista Ciência Agronômica is the property of Revista Ciencia Agronomica and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

As artificial neural networks (ANN), they are computational models inspired by the way the nervous system of living beings work, these models can be used for processing and classification of data and applications, such as series and function prediction. Thus, this work used a time-delayed neural network (TDNN) to predict the demand for active energy on the P4 bus in the city of Presidente Prudente. [ABSTRACT FROM AUTHOR]

Copyright of Ingeniería Solidaria is the property of Universidad Cooperativa de Colombia and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Revista Gestão & Tecnologia is the property of Revista Gestao & Tecnologia and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Today's globalized and highly competitive environment imposes chicken producers two major challenges: resource optimization and cost reduction. In this context, this work aims to use artificial neural network models and multiple linear regression for analysis and prediction of broiler productive variables of an agribusiness of Paraná. The applicability of recurrent neural networks to predict the price of kilo of frozen and chilled chicken is also analyzed. The database provided by the company has a 2-year history of movement, containing the main production variables, for 4500 batches of Coob, Coob Fast and Coob Slow strains. In the analysis of the applicability of recurrent neural networks we used a 11-year database provided by the Center for Advanced Studies in Applied Economics (CEPEA). The results show that the forecast models provide reliable estimates for the response variables: Average Weight and Productive Efficiency Ratio and demonstrate the effectiveness of the recurring LSTM predictions for the price of kilo of frozen and chilled chicken for a short term horizon. [ABSTRACT FROM AUTHOR]

Copyright of Revista de Gestão, Finanças E Contabilidade is the property of Universidade do Estado da Bahia and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

O gerenciamento de recarga de combustível em um reator de potência tem como principal foco o padrão de recarga no núcleo de modo a alcançar melhor rendimento no ciclo observando todos os parâmetros de segurança adotados. Neste artigo apresentamos a estratégia de um algoritmo baseado em redes neurais artificiais que poderia ser usado para otimizar a recarga de combustível num reator nuclear. A ideia é apresentar uma metodologia baseada em rede neural feedforward multicamadas baseada em neurônios multi-valorados, que poderá ser usada para desenvolver uma metodologia capaz de escolher as melhores combinações que satisfaçam o fator de pico de potência radial e maximizem o fator de multiplicação efetivo no início do ciclo, e também satisfaçam a relação de potência crítica mínima e taxa máxima de geração de calor no final do ciclo.

Copyright of Ciência Florestal (01039954) is the property of Ciencia Florestal and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The following paper presents the main results of exploratory research oriented to the design and implementation of a pattern recognition system for flatfoot identification in children between 5 and 6 years. Patterns were determined from texture analysis of foot thermographic images, and from contour analysis of footprint images. For each case, an artificial neuronal network was trained, with base in a back-propagation algorithm. In each trial, 70% of data were used for training, and 30% for validation. For experiments done, success rates greater than 80% were achieved. The best results were reached with contour patterns reduced by principal components analysis, PCA, in a binary system, with a success rate of 90.84% in crossvalidation. Results are a contribution to the study of diagnostic techniques for flatfoot treatment through the use of technologic tools. [ABSTRACT FROM AUTHOR]

Copyright of Nativa is the property of Revista Nativa and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Any atypical change in a procedure can be characterized as a "failure". Consequently, it may result in economic losses and/or a rise of the operational cost, because most of the times the process will need to be interrupted. Therefore, the concern with the quality and security of the processes has stimulating studies of diagnosis and monitoring failures in industrial equipments. In light of this, the present article has as purpose to apply three different methods (Artificial Neural Networks - ANN, Fuzzy Logic - FL and Support Vector Machine - SVM). All of those were applied as detection and classification systems of failure in the processes of a case study in order to diagnose these artificial intelligence techniques so that the efficiency of each method can be compared. All investigation is done by modeling a reactor of Van der Vusse's kinetic causing four types of failures, in the concentration of a reagent (failure 1), in the sensor which measures the concentration of the interested product and temperature (failure 2 and 3), and in the valve locking (failure 4). The data used in this methodology is based in quantitative and qualitative historical information. All methods are able to detect failures, but in different times. ANN is the one which detects faster all the failures. SVM detects some minutes later, however with good precision, even though this method uses less computational effort compared to ANN. Fuzzy, in the most of the cases studied, takes hours to detect any change in the system, which makes this one the less effective among the ones studied. [ABSTRACT FROM AUTHOR]

Estimates of tree bark thickness are fundamental for forest management, however, the degree of precision is conditioned to the adoption of efficient modeling techniques. The objective of this study was to evaluate and propose a model of artificial neural networks to estimate the thickness of the tree bark of Tectona grandis (Teak). The data originated from the measurement of 68 dominant trees, ranging in age from 6 to 33 years. The thickness of the bark was correlated with variables inherent to the tree, being: diameter in the different positions of the stem (di); diameter at 1.3 m height (dbh); total height (ht); relative height (hi_rel); and age (id). The trained networks were of the multilayer perceptron type, and a linear regression model was adjusted as a comparative support. The accuracy of the estimates was evaluated through statistical indicators and graphical analysis. The results showed a strong correlation between bark thickness and tree diameter, as well as relative height, with values above 0.70. Age also exerted a strong influence on the thickness of the bark of the trees. The artificial intelligence technique has demonstrated the potential for such application and the model proposed with the input variables: diameter, relative height and age was the one that presented the best statistical performance, and thus was the most suitable for predicting the bark in Teak trees. [ABSTRACT FROM AUTHOR]

The classification of supernovae (explosions of certain stars) divides them into two main types, those of type I do not present Hydrogen in the spectrum while those of type II present. In addition to the division into these two types, there is still a subdivision that establishes types Ia, Ib and Ic. In practice, the classification of supernovae requires specialized knowledge of astronomers and data (light spectra) of good quality. Some automatic/intelligent classifiers have been developed and are reported in the literature, one of them is CIntIa, which uses 4 Artificial Neural Networks to classify supernovae types Ia, Ib, Ic and II. The objective of this work is to improve CIntIa, so that it has more diversity in its learning, proposing CIntIa 2.0. In this way, this work is a hierarchical learning structure that connects Artificial Neural Networks in an integrated system that allows a more secure and unambiguous classification. The computational improvement of this new version included the increased amount of data used at all stages of development of intelligent classifier and a new approach to filtering and processing of spectral data, ensuring better quality of information that are to be trained networks. The results achieved were good, especially in the classification of types Ia and II. A comparison with the works found in the literature shows that CIntIa 2.0 is superior in quantity and diversity of data and achieves higher classification indices than the other classifiers. [ABSTRACT FROM AUTHOR]

Nowadays, the demand forecasting has become a challenge to stock managers due to the need to balance the service from the supply chain through the administration of the availability of the materials and investment in stock. This study suggests an alternative method to the demand forecasting of the ferroalloy used in the production of flat steel in Companhia Siderúrgica Nacional (CSN) through a computer code based on Artificial Neural Networks (ANN). The methodology consists in implementing the main elements that model the neural networks, such as: selection of variables, number of variables, network architecture, training and outflow forecasting. The model was developed using the Multi-Layer Perceptron (MLP) concept, with a data set such as: specific intake, lead time, monthly and annual production of steel mills, safety stock, from 2014 to 2016. The obtained results with the use of the ANNs were compared to the values measured during the period above and also to a previously developed model based on multiple linear regression. The results obtained through the ANNs showed a higher of predictability compared to the method of multiple linear regression. [ABSTRACT FROM AUTHOR]

Copyright of Revista Iberoamericana de Ingeniería Mecánica is the property of Editorial UNED and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Perspectivas em Ciência da Informaçao is the property of Nova Economia and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Revista Gestão & Tecnologia is the property of Revista Gestao & Tecnologia and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Artificial Intelligence (AI) can be understood as a field of study in the, through its own tools, has the purpose of developing systems and "intelligent" softwares, capable of optimizing human performance in several areas of knowledge, including biomedical. In the gastroenterology, hundreds of papers have addressed the topic and its applications in the last 10 years, revealing the need for a review study on the subject. In the present study, 1146 publications were collected and 125 were selected for analysis. Most of the studies used Artificial Neural Networks (ANN) as the main AI tool, followed by Supporting Vector Machine (SVM) and Machine Learning. Among the purposes of the use of AI techniques, the (i) classification, used for differentiation of patients and non-patients, tumors, carcinomas and other information related to medical diagnosis and (ii) prediction, involving patient survival time or the prediction of development of metastases, for example. Among the various gastroenterological diseases, the work in the area of oncology (colorectal, gastric, pancreatic and hepatic cancer) predominated. The most used inputs to the system were imaging and genetic data, in addition to others, in order to optimize the diagnosis, prognosis and prediction of response to therapy in the most important affections of the specialty. [ABSTRACT FROM AUTHOR]